And C-Suites are freaking the fuck out (my company) because we’re completing all of the work in <40 hours per week, which causes them to lose money lmao It is so fucking wild to me how no one sees the glaring problem here

All of this AI shit should’ve stayed as nothing more and nothing less than a tool to make the very monotonous work done by everyone just a tad easier.

It feels like we’re in such a massive fucking bubble right now on the scale of how people have talked about the sub prime mortgage crisis.

My boomer ass parents just keep telling me to wait for things to get better, but how can you wait for things to get better when things are just statistically worse for everyone

  • rah@feddit.uk
    link
    fedilink
    English
    arrow-up
    23
    ·
    7 months ago

    we’re completing all of the work in <40 hours per week

    And this is a change due to using an LLM? Assuming so, I’m absolutely astonished. Can I ask what work the quicker workers do, and how an LLM has helped them to be quicker?

      • abc [he/him, comrade/them]@hexbear.net
        link
        fedilink
        English
        arrow-up
        12
        ·
        7 months ago

        a lot of my job is ‘emails’ but I can’t just use a LLM for responses - it’d be so obviously bad and hallucinate solutions to problems that don’t exist. god I wish.

        • impartial_fanboy [he/him]@hexbear.net
          link
          fedilink
          English
          arrow-up
          5
          ·
          7 months ago

          If you have enough emails from your job you could fine-tune one of the models on them and it would probably do pretty okay. It’d be really cheap cus it’s not that much data, assuming you don’t have the hardware to do it yourself.

          • ped_xing [he/him]@hexbear.net
            link
            fedilink
            English
            arrow-up
            4
            ·
            7 months ago

            It would sound like me, sure. It would generate emails saying that a course of action will or won’t succeed. The main difference would be that my emails are informed by investigating things often not captured in the email chain to which I’m replying. robo-ped_xing would be just as confident as me, much cheaper, faster and utterly wrong very often. If that’s an OK trade-off, they could have replaced me with a magic 8-ball years ago.

          • abc [he/him, comrade/them]@hexbear.net
            link
            fedilink
            English
            arrow-up
            2
            ·
            7 months ago

            Yeah, I could easily train a model on the multiple evernote notebooks I’ve accumulated over the years but to be honest it’d just basically replace me copying & pasting from Evernote and not actually save me any significant time on the brunt of my job - which is usually solving new and unique problems via email that the software engineers cause with every release.

      • rah@feddit.uk
        link
        fedilink
        English
        arrow-up
        1
        ·
        edit-2
        7 months ago

        Thanks for your response. Alas, I’m still completely oblivious as to how ChatGPT helps you work quicker.